If I remember right they use the equivalent of a ledger of changes to manage volume state. So in this case, they copy over the contents (up to a certain point i

If I remember right they use the equivalent of a ledger of changes to manage volume state. So in this case, they copy over the contents (up to a certain point in time) to the new faster virtual volume, then append and direct all new changes to the new volume. This is also how they are able to snapshot a volume at a certain point in time without having any downtime or data inconsistencies.